Small Target Detection Algorithm Based on Improved YOLOv8

YOLOv8 algorithm is a deep learning method, which has achieved quite excellent performance in target detection tasks, but the detection effect of small targets and dense targets is not ideal. The adaptability of targets with different scales and shapes is not good enough, and the model generalization ability needs to be improved. In order to solve these problems, the YOLOv8 algorithm is improved from two aspects: adaptive image scaling and CA. The adaptive image scaling algorithm based on edge operator improves the accuracy of edge and interpolation direction detection, has better edge estimation and interpolation effect, and improves the visual quality of the scaled image. CA integrates location information into channel attention and aggregates features along two spatial directions, avoiding a lot of computational overhead. The results show that the improved model proposed in this paper combines all modules and strategies together and has better performance of small target detection.